[xcp] Notes from XCP meeting at ISI (2/25/04)

[Aaron Falk]

Aman and I met today.  He has finished updating the code to use 
congestion headers with throughput, rather than cwnd.  We briefly 
discussed what to do next.

In the near term, Aman is going to work on the following tasks:

0. Fix sender code to use local RTT when converting reverse_feedback 
(throughput) to cwnd

1. Plot BTA & BTF to check that the residual feedback pools are doing 
the right things

2. Plot avg_rtt to check convergence dynamics

3. With existing testbed configuration, run concurrent flows with 
different RTTs.

4. Change the desired_throughput of some flows to be less than the fair 
allocation, check that the router does the right thing and makes use of 
the extra capacity.

5. Fix router code to support multiple instances.

6. Add a second router in series, run cross traffic:

           -----------R1
          /
S2=====X1======X2====R2
       /
S1---/

All links are 100Mbps.
Cross-traffic from S1 goes to R1
Traffic of interest goes from S2 to R2
Create a bottleneck in X1 with cross traffic, then remove it.